Decided that I’m going to go with “collaborative diffusion” pathfinding–I created a test file that implements this system of path finding quickly and easily.
Well, somewhat. The current file set I’ve got up there calculates a path from the bottom left of the screen to the middle 30 times, and it takes ~.43 seconds on my machine to do that. I’m going overly generous in terms of the depth parameter–reducing that cuts the time down significantly–and my AI won’t need the full path to the target anyway using this method, so in terms of processing time I don’t think it’ll be an issue. Dijkstra’s algorithm would’ve been nice, as would A*, with the latter I hear being the fastest, but I think with some playing I can get what I’ve got now to suit my needs.
*Oh, and I discovered switching to numpy was about as simple as doing find/replace on all “Numeric” entries with “numpy”. Heh.
Some more testing shows that the floodfill part of this is the most expensive part of the algorithm. Should’ve been obvious actually, but, eh.